Multi-target tracking method and device of millimeter wave radar, and storage medium

Abstract

The invention discloses a multi-target tracking method and device for a millimeter-wave radar and a storage medium, relates to the technical field of radar signals, and discloses the multi-target tracking method and device for the millimeter-wave radar and the storage medium, and the method comprises the steps: carrying out the multi-frame accumulation processing of a radar signal through a sliding window, the condition that radar signals are missing at individual moments due to the flicker problem existing in the radar signals is avoided, meanwhile, effective point cloud data are extracted from a to-be-matched point cloud set through an association gate to group an initial confirmation matrix, so that the scene corresponding to each radar signal is partitioned to reduce the dimensionality of the initial confirmation matrix; the target confirmation matrix is obtained, joint probability data association is executed based on the target confirmation matrix, the number of interconnection matrixes obtained based on splitting of the initial confirmation matrix is reduced, and the change relation of the number of joint events corresponding to the interconnection matrixes along with the to-be-matched point cloud data is converted into an approximately linear relation; and the computing resources of the joint probability data association algorithm are greatly reduced.

Description

Technical Field

[0001] The present application relates to the field of radar signal technology, and in particular to a multi-target tracking method, device and storage medium for millimeter-wave radar. Background Art

[0002] Target tracking is the core technology of radar signal processing, and plays a key role in intelligent transportation systems. It can predict and update the vehicle's motion state and trajectory by analyzing the target information detected by the millimeter-wave radar, such as distance, speed, azimuth, etc., so as to realize vehicle obstacle detection and avoidance, adaptive cruise and other functions.

[0003] In the field of millimeter wave radar tracking, the most widely used multi-target tracking algorithms are based on Bayesian theory, which include key components such as motion model, track initiation, data association and state filtering. Among them, data association is the core issue in multi-target tracking algorithms. Classic data association algorithms include joint probability data association algorithm (JPDA, Joint Probability Data Association), etc. However, the number of joint events in the JPDA algorithm is an exponential function of all candidate measurements, which increases rapidly with the increase of measurement density, resulting in an explosion in calculation volume and a huge memory burden. Therefore, the JPDA algorithm is difficult to apply to engineering implementation. Summary of the invention

[0004] The main purpose of this application is to provide a multi-target tracking method, device and storage medium for millimeter-wave radar, aiming to solve the technical problem of excessive computing resources in the current multi-target tracking based on the joint probabilistic data association algorithm.

[0005] To achieve the above object, the present application proposes a multi-target tracking method of a millimeter wave radar, and the multi-target tracking method of the millimeter wave radar includes:

[0006] By setting multiple sliding windows, the radar signals are processed with multi-frame accumulation and density-based clustering to obtain the point cloud set to be matched, and the predicted point cloud set is calculated by traversing the target point cloud set at the previous moment;

[0007] Establishing an association gate according to the association relationship between the predicted point cloud set and the to-be-matched point cloud set;

[0008] Based on the association gate, valid point cloud data in the to-be-matched point cloud set and first point cloud data corresponding to the valid point cloud data in the predicted point cloud set are acquired, and an initial confirmation matrix is ​​established according to the association relationship between the valid point cloud data and the first point cloud data;

[0009] Acquire multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix, group the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index into one group, obtain multiple groups, and establish a target confirmation matrix belonging to the groups, perform joint probability data association based on the target confirmation matrices of each group, wherein the predicted point cloud set includes several predicted point cloud data.

[0010] In one embodiment, the step of setting multiple sliding windows, performing multi-frame accumulation processing and density-based clustering processing on the received radar signals, and obtaining a point cloud set to be matched includes:

[0011] Setting a plurality of sliding windows, storing a plurality of radar signals sequentially output by the millimeter-wave radar in corresponding sliding windows, and extracting radar signals stored in a plurality of frames of the sliding windows when a sliding window operation is detected to be completed, to obtain a plurality of frames of point cloud data;

[0012] The multi-frame point cloud data is subjected to density-based clustering processing to obtain the point cloud set to be matched.

[0013] In one embodiment, the millimeter wave radar includes a plurality of target trackers, one of the target trackers is used to track a point cloud data, and the step of traversing the point cloud set to be matched and calculating the predicted point cloud set includes:

[0014] Initializing the target tracker;

[0015] After traversing the target track corresponding to each target point cloud data in the target point cloud set at the previous moment, the motion state prediction value of each target point cloud data at the current moment and the state covariance matrix prediction value at the current moment are calculated through the uniform acceleration motion model to obtain the predicted point cloud set.

[0016] In one embodiment, the to-be-matched point cloud set includes a plurality of to-be-matched point cloud data, and the step of establishing an association gate according to the association relationship between the predicted point cloud set and the to-be-matched point cloud set includes:

[0017] Traversing each of the to-be-matched point cloud data, and traversing the predicted point cloud data corresponding to the target point cloud data based on the traversed to-be-matched point cloud data, to obtain target predicted point cloud data corresponding to each of the target point cloud data;

[0018] According to the point cloud data to be matched, the association gate is established on each of the target prediction point cloud data.

[0019] In one embodiment, the step of acquiring valid point cloud data in the to-be-matched point cloud set based on the association gate comprises:

[0020] By means of the association gate, determining whether the target point cloud data falls within the association range of the association gate;

[0021] If the target point cloud data falls within the associated range, the to-be-matched point cloud data corresponding to the target point cloud data falling within the associated range is extracted as the valid point cloud data.

[0022] In one embodiment, the step of performing joint probabilistic data association according to the target confirmation matrix of each of the groups comprises:

[0023] Traversing each of the target confirmation matrices, and splitting each of the target confirmation matrices to obtain a plurality of interconnection matrices, wherein each of the interconnection matrices represents a feasible joint event;

[0024] Calculating the probability of the feasible joint event according to the interconnection matrix to obtain first probability data, and calculating the probability that the valid point cloud data included in the interconnection matrix belongs to the first point cloud data to obtain second probability data;

[0025] According to the comparison relationship between the second probability data and the second preset probability threshold, if the second probability data is greater than the second preset probability threshold, the matching state of the valid point cloud data corresponding to the second probability data is marked as a correct matching state, wherein the first point cloud data in the correct matching state is determined as the first point cloud data to be calculated, and the valid point cloud data in the correct matching state is determined as the second point cloud data to be calculated;

[0026] Calculate the Kalman gain and the new information vector of the first point cloud data to be calculated under the uniform acceleration motion model, and update the motion state prediction value and the state covariance matrix prediction value according to the Kalman gain and the new information vector to obtain the motion state and the posterior state covariance of the first point cloud data to be calculated.

[0027] In one embodiment, after the step of marking the matching state of the valid point cloud data corresponding to the second probability data as a correct matching state if the second probability data is greater than the second preset probability threshold, the method further includes:

[0028] Determine the valid point cloud data other than the second point cloud data to be calculated as the point cloud data to be determined whose matching status is an incorrect matching status;

[0029] Determine whether the point cloud data to be determined meets the preset creation conditions;

[0030] If the point cloud data to be determined meets the preset creation condition, the point cloud data to be determined is determined as the first next target point cloud data of the next target tracking process.

[0031] In one embodiment, after the step of performing joint probabilistic data association according to the target confirmation matrix, the method further comprises:

[0032] Traversing the tracks and point cloud attributes corresponding to each of the first point cloud data, extracting the first point cloud data with a track status of a death track from each of the tracks, and extracting the first point cloud data with an attribute status of an abnormal attribute from each of the point cloud attributes, then eliminating the first point cloud data with a track status of the death track and / or the first point cloud data with an attribute status of an abnormal attribute, to obtain the second next target point cloud data;

[0033] The first next target point cloud data and the second next target point cloud data are passed into the next target tracking process.

[0034] In addition, to achieve the above-mentioned objectives, the present application also proposes a multi-target tracking device of a millimeter-wave radar, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the multi-target tracking method of the millimeter-wave radar as described above.

[0035] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the multi-target tracking method of the millimeter-wave radar as described above are implemented.

[0036] One or more technical solutions proposed in this application have at least the following technical effects:

[0037] By setting multiple sliding windows, multi-frame accumulation processing and density-based clustering processing are performed on the accessed radar signals to obtain a point cloud set to be matched, and the predicted point cloud set is calculated by traversing the point cloud set to be matched; an association gate is established based on the association relationship between the predicted point cloud set and the point cloud set to be matched; based on the association gate, valid point cloud data in the point cloud set to be matched and the first point cloud data corresponding to the valid point cloud data in the predicted point cloud set are obtained, and an initial confirmation matrix is ​​established according to the association relationship between the valid point cloud data and the first point cloud data; multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix are obtained, and the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index are grouped into one group to obtain multiple groups, and a target confirmation matrix belonging to the group is established, and joint probability data association is performed based on the target confirmation matrix of each group, wherein the predicted point cloud set includes a number of predicted point cloud data.

[0038] The present application performs multi-frame accumulation processing on radar signals by setting a sliding window, and performs position compensation on the sliding window for the radar signals connected at the previous moment according to their motion state, thereby avoiding the missing of radar signals at individual moments caused by the flickering problem of the radar signal. At the same time, it is proposed to extract valid point cloud data from the point cloud to be matched through an association gate, establish an initial confirmation matrix through the association relationship, and group the initial confirmation matrix, so as to partition the scenes corresponding to each radar signal to achieve the reduction of the dimension of the initial confirmation matrix, obtain multiple low-dimensional target confirmation matrices, and perform joint probability data association based on the low-dimensional target confirmation matrix, thereby reducing the number of interconnected matrices obtained by splitting the initial target confirmation matrix, and changing the exponential growth relationship between the number of joint events corresponding to the original interconnected matrix and the number of target point cloud data into an approximately linear relationship, thereby greatly reducing the computing resources of the joint probability data association algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0041] Figure 1 A flowchart of a first embodiment of a multi-target tracking method for a millimeter-wave radar of the present application is provided;

[0042] Figure 2This is a schematic diagram of the overall process of the multi-target tracking method of the millimeter wave radar in this application;

[0043] Figure 3 A flowchart diagram of a second embodiment of a multi-target tracking method for a millimeter-wave radar of the present application;

[0044] Figure 4 A flowchart diagram of a third embodiment of the multi-target tracking method of the millimeter-wave radar of the present application;

[0045] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the multi-target tracking method of millimeter-wave radar in the embodiment of the present application.

[0046] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0047] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0048] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0049] The main solution of the embodiment of the present application is: by setting multiple sliding windows, multi-frame accumulation processing and density-based clustering processing are performed on the accessed radar signals to obtain a point cloud set to be matched, and by traversing the point cloud set to be matched, a predicted point cloud set is calculated; an association gate is established based on the association relationship between the predicted point cloud set and the point cloud set to be matched; based on the association gate, valid point cloud data in the point cloud set to be matched and the first point cloud data corresponding to the valid point cloud data in the predicted point cloud set are obtained, and an initial confirmation matrix is ​​established according to the association relationship between the valid point cloud data and the first point cloud data; multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix are obtained, and the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index are grouped into a group to obtain multiple groups, and a target confirmation matrix belonging to the group is established, and joint probability data association is performed based on the target confirmation matrix of each group, wherein the predicted point cloud set includes a number of predicted point cloud data.

[0050] In the field of millimeter-wave radar tracking, the most widely used multi-target tracking algorithms are based on Bayesian theory. These algorithms include key components such as motion model, track initiation, data association and state filtering. Among them, data association is the core issue in multi-target tracking algorithms. Classic data association algorithms include joint probability data association algorithm (JPDA) and others. In the JPDA algorithm, the number of joint events is an exponential function of all candidate point cloud data. It increases rapidly with the increase of point cloud data density, resulting in an explosion in computational complexity and a huge memory burden. Therefore, the JPDA algorithm is difficult to apply to engineering projects. At present, most of the multi-target tracking algorithm research based on JPDA focuses on increasing the restrictions of elliptical wave gates to reduce the number of executable joint events. The threshold conditions need to be set based on experience, and the relationship between the exponential growth of the number of joint events and the number of point cloud data has not been fundamentally changed.

[0051] The present application provides a solution, which performs multi-frame accumulation processing on radar signals by setting a sliding window, and performs position compensation on the sliding window for the radar signals connected at the previous moment according to their motion state, thereby avoiding the missing of radar signals at individual moments caused by the flickering problem of the radar signal. At the same time, it is proposed to extract effective point cloud data from the to-be-matched point cloud set through an association gate to group the initial confirmation matrix, so as to partition the scenes corresponding to each radar signal to achieve the reduction of the dimension of the initial confirmation matrix, obtain the target confirmation matrix, and perform joint probability data association based on the target confirmation matrix, thereby reducing the number of interconnected matrices obtained based on the splitting of the initial confirmation matrix, and changing the exponential growth relationship between the number of joint events corresponding to the original interconnected matrix and the number of to-be-matched point cloud data into an approximately linear relationship, thereby greatly reducing the computing resources of the joint probability data association algorithm.

[0052] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a multi-target tracking device of a millimeter-wave radar, etc. The following takes the multi-target tracking device of a millimeter-wave radar as an example to illustrate this embodiment and the following embodiments.

[0053] Based on this, the embodiment of the present application provides a multi-target tracking method of millimeter wave radar, referring to Figure 1 Combined with Figure 2 , Figure 1 This is a flow chart of the first embodiment of the multi-target tracking method of the millimeter-wave radar of the present application.

[0054] In this embodiment, the multi-target tracking method of the millimeter wave radar includes steps S10 to S40:

[0055] Step S10, by setting multiple sliding windows, multi-frame accumulation processing and density-based clustering processing are performed on the connected radar signals to obtain a point cloud set to be matched, and by traversing the target point cloud set at the previous moment, a predicted point cloud set is calculated.

[0056] Because the point cloud data corresponding to the radar signal transmitted by the millimeter-wave radar has a flickering problem, this flickering problem will cause the point cloud data at certain moments to be missing. In a cluttered environment, the missing point cloud data will reduce the robustness of multi-target tracking to a certain extent.

[0057] Therefore, in order to avoid the above problems, the present embodiment proposes to first set up multiple sliding windows, and after storing the accessed radar information through the sliding windows, perform multi-frame accumulation processing on the radar signals stored in the sliding windows. When a radar signal from a millimeter-wave radar is received at the current moment, the radar information is stored at the front end of the sliding window. When entering the next moment and receiving a new radar signal from the millimeter-wave radar at the next moment, the position of the radar signal stored in the sliding window at the first few moments of the next moment is changed. Specifically, position compensation is performed according to the time interval with the new radar signal and its corresponding motion state, thereby completing a sliding window operation.

[0058] The multi-frame radar signals corresponding to a sliding window operation are clustered based on density, so as to cluster the point cloud data corresponding to the radar signals with similar characteristics, and obtain the point cloud set to be matched. The specific steps are shown in steps S13 to S14. Then, the target point cloud set at the previous moment is traversed and calculated to obtain the predicted point cloud set.

[0059] In a feasible implementation manner, step S10 includes steps S11 to S12:

[0060] Step S11, initializing the target tracker;

[0061] It should be noted that the millimeter wave radar includes multiple target trackers, one target tracker is used to track one radar signal (ie, point cloud data), and the following description is based on one target tracker.

[0062] If it is determined that the target tracker has not been initialized, the state transfer matrix (F) of the target tracker, the covariance matrix (R) of the observation noise, and the covariance matrix (Q) of the process noise of the system state transfer model are initialized.

[0063] Step S12, after traversing the target track corresponding to each target point cloud data in the target point cloud set at the previous moment, the motion state prediction value of each target point cloud data at the current moment and the state covariance matrix prediction value at the current moment are calculated through the uniform acceleration motion model to obtain the predicted point cloud set (i.e. Figure 2 Step S6 in the above process).

[0064] One target point cloud data corresponds to a target track of the vehicle. At this time, the target track corresponding to each target point cloud data in the target point cloud set obtained at the previous moment is traversed. Based on the uniform acceleration motion model, the x, y, and v of the target track input to the Kalman filter are x 、v y 、a x and a y Six motion states are predicted (i.e. Figure 2 Step S5), specifically, based on expression ①.

[0065] X t (KK-1)=FX t (K-1)————Expression ①

[0066] And, the state covariance matrix of the target track input to the Kalman filter is predicted, specifically, based on expression ②.

[0067] P t (KK-1) = FP t (K-1)F T +Q————Expression②

[0068] Among them, K represents the time, t represents the t-th target track, and X t (KK-1) represents the motion state prediction value of the target track at time K after the information at time K-1 is updated (i.e., the motion state prediction value at the current time), P t (KK-1) represents the predicted value of the state covariance matrix at time K after the information at time K-1 is updated (i.e., the predicted value of the state covariance matrix at the current time).

[0069] Step S20: establishing an association gate (i.e. Figure 2 Step S7 in ).

[0070] In order to screen the point cloud data to be matched included in the point cloud set to be matched and obtain valid point cloud data associated with any predicted point cloud data in the predicted point cloud set, this embodiment establishes an association gate to determine the association between the predicted point cloud data and its corresponding point cloud data to be matched.

[0071] It should be noted that the point cloud data to be matched is the motion state point cloud data of a target actually acquired at a new moment, and the predicted point cloud data is the point cloud data of the target point cloud data at the previous moment predicted according to the motion state. Therefore, the point cloud data to be matched that actually corresponds to the same target at the same moment has a corresponding relationship with the predicted point cloud data predicted for the same target at the same moment.

[0072] In a feasible implementation manner, step S20 includes steps S21 to S22:

[0073] Step S21, traversing each of the to-be-matched point cloud data, and traversing the predicted point cloud data corresponding to the target point cloud data according to the traversed to-be-matched point cloud data, to obtain target predicted point cloud data corresponding to each of the target point cloud data;

[0074] Step S22: establishing the association gate on each of the target prediction point cloud data according to the point cloud data to be matched.

[0075] Specifically, by traversing each to-be-matched point cloud data in the to-be-matched point cloud set, and in the traversal process, traversing the predicted point cloud data corresponding to the traversed to-be-matched point cloud data, after obtaining the predicted point cloud data corresponding to each to-be-matched point cloud data, the specific expressions ③ to ⑥ are used to establish an association gate on the predicted point cloud data (i.e. Figure 2 Step S7 in the process).

[0076] Z t (KK-1) = HX t (KK-1)————Expression ③

[0077]

[0078] S t (K)=HP t (K)H T +R————Expression ⑥

[0079] H represents the observation matrix, Z t (KK-1) represents the predicted value of the observation at time K after the information at time K-1 is updated. represents the predicted new information vector of the jth valid point cloud data and the tth predicted point cloud data, z j (K) represents the measured value of the jth valid point cloud data, γ is the correlation range of the correlation gate, S t (K) is the observation error covariance matrix.

[0080] The observation matrix of the predicted point cloud data is input into expression ⑥, and the observation error covariance matrix of the predicted point cloud data in the uniform acceleration motion model is output. At the same time, according to the predicted point cloud data at the K-1 moment, the observation prediction value of the predicted point cloud data at the K moment is updated by expression ③ using the uniform acceleration motion model. Based on the observation prediction value, the predicted point cloud data and the predicted point cloud data to be matched corresponding to the predicted point cloud data are obtained by expression ④. The predicted new information vector and the corresponding observation error covariance matrix are substituted into expression ⑤ to obtain the judgment value d of the target point cloud data. 2 , determine whether the judgment value is within the valid range γ. If the judgment value is within the valid range, it means that the target point cloud data matches its associated point cloud data to be matched.

[0081] Step S30, based on the association gate, obtain the valid point cloud data in the to-be-matched point cloud set and the first point cloud data corresponding to the valid point cloud data in the predicted point cloud set, and establish an initial confirmation matrix (i.e. Figure 2 Step S9 in the process).

[0082] Through the association gate of the above step S20, if the judgment value corresponding to the point cloud data to be matched is within the valid range, the point cloud data to be matched is determined as valid point cloud data. At the same time, because the predicted point cloud data is the predicted data of the point cloud data to be matched at a new moment, the predicted point cloud data at the same moment has a corresponding relationship with the point cloud data to be matched. At this time, the predicted point cloud data corresponding to the valid point cloud data is obtained, and after the predicted point cloud data is determined as the first point cloud data, an initial confirmation matrix is ​​established according to the association relationship between the valid point cloud data and the first point cloud data.

[0083] In a feasible implementation manner, step S30 includes steps S31 to S32:

[0084] Step S31, through the association gate, determine whether the target point cloud data falls within the association range of the association gate (i.e. Figure 2 Step S8 in the process).

[0085] Step S32: if the target point cloud data falls within the associated range, the to-be-matched point cloud data corresponding to the target point cloud data falling within the associated range is extracted as the valid point cloud data.

[0086] It can be seen from expressions ③ to ⑥ that the association gate is used to determine whether the corresponding target point cloud data is within the association range of the association gate, that is, whether the judgment value corresponding to the target point cloud data is within the valid range. If the point cloud data to be matched is within the association range, the point cloud data to be matched is determined as valid point cloud data.

[0087] If the point cloud data to be matched is not within the associated range, it means that the target motion state corresponding to the point cloud data to be matched does not match the motion state of the current target predicted point cloud data, and there is an abnormal motion state. Therefore, the point cloud data to be matched is determined as abnormal point cloud data and is not included in the execution calculation of subsequent process steps.

[0088] Step S40, obtaining multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix, grouping the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index into one group, obtaining multiple groups, and establishing a target confirmation matrix belonging to the group (i.e. Figure 2 In step S10), joint probability data association is performed according to the target confirmation matrix of each of the groups, wherein the predicted point cloud set includes a plurality of predicted point cloud data.

[0089] First, according to each valid point cloud data in each initial confirmation matrix, after confirming multiple predicted point cloud indexes associated with each valid point cloud data, each valid point cloud data is traversed. If the traversal obtains that there is a common predicted point cloud index between two valid point cloud data, the two valid point cloud data are grouped into one group. The predicted point cloud index set of this group is the union of the common predicted point cloud indexes between the two valid point cloud data. Then, other valid point cloud data are traversed to confirm whether there is any valid point cloud data belonging to the same predicted point cloud index set as the predicted point cloud index set of this group. If so, the valid point cloud data is grouped into this group, and so on, until all valid point cloud data are divided into multiple groups, and the target confirmation matrices of each group are respectively confirmed according to the multiple groups obtained. The valid point cloud data in each target confirmation matrix are connected with other valid point cloud data in the target confirmation matrix through the predicted point cloud index commonly associated with it. Figure 2 A Group in is a group.

[0090] Explain with reference to expression ⑦.

[0091]

[0092] Among them, w jt represents the target confirmation matrix with j rows and t columns, n represents the group index to which the target confirmation matrix belongs, and expression ⑦ determines whether the valid point cloud data in the target confirmation matrix has an association relationship with the first point cloud data, specifically:

[0093] In the process of traversing each valid point cloud data, it is determined whether the first point cloud data corresponding to the traversed valid point cloud data belongs to the predicted point cloud index. If so, the output of expression ⑦ is 1; if not, the output of expression ⑦ is 0.

[0094] The target confirmation matrix finally obtained after grouping only establishes the association relationship between the valid point cloud data in the group and the predicted point cloud data in the group, wherein the first column in the target confirmation matrix indicates that the valid point cloud data may come from clutter.

[0095] The number of interconnected matrices obtained by splitting the target confirmation matrix after grouping is reduced, and the relationship between the number of joint events corresponding to the interconnected matrix and the exponential growth of the number of point cloud data to be matched becomes an approximately linear relationship, which greatly reduces the computing resources of the joint probability data association algorithm, as described in steps S41 to S44.

[0096] In a feasible implementation manner, step S40 includes steps S41 to S44:

[0097] Step S41, traverse each of the target confirmation matrices, and split each of the target confirmation matrices to obtain multiple interconnected matrices (i.e. Figure 2 In step S11), each of the interconnection matrices represents a feasible joint event.

[0098] Traverse each target confirmation matrix to split each target confirmation matrix to obtain an interconnection matrix. Each interconnection matrix represents a feasible interconnection event θ i ,i=1,2,...,L, where L is the sum of feasible joint events.

[0099] Step S42, calculating the probability of the feasible joint event according to the interconnection matrix to obtain first probability data, and calculating the probability that the valid point cloud data included in the interconnection matrix belongs to the first point cloud data to obtain second probability data (i.e. Figure 2 , step S12 in the process.

[0100] The probability of a feasible joint event is calculated based on expressions ⑧~⑨.

[0101]

[0102] Where T represents the number of valid point cloud data in the feasible joint event, Z K represents the predicted value of the motion state at time K, P D Represents the probability of correct detection in a valid point cloud, FM represents the number of valid point cloud data that do not match the predicted point cloud index in the executable joint event, Represents the new information vector of the valid point cloud data and the first point cloud data associated with it, S tj represents the observation covariance matrix, M represents the dimension of the new information vector, P n {θ i Z K} represents the first probability data.

[0103] After the first probability data is calculated based on expressions ⑧ to ⑨, the second probability data is calculated based on expression ⑩.

[0104]

[0105] Among them, β jt The second probability data indicating that the j-th valid point cloud data belongs to the t-th first point cloud data.

[0106] Step S43, based on the comparison relationship between the second probability data and the second preset probability threshold, if the second probability data is greater than the second preset probability threshold, the matching status of the valid point cloud data corresponding to the second probability data is marked as a correct matching status, wherein the first point cloud data in the correct matching status is determined as the first point cloud data to be calculated, and the valid point cloud data in the correct matching status is determined as the second point cloud data to be calculated.

[0107] Step S44, calculate the Kalman gain and the innovation vector of the first point cloud data to be calculated under the uniform acceleration motion model, and update the motion state prediction value and the state covariance matrix prediction value according to the Kalman gain and the innovation vector to obtain the motion state and the posterior state covariance matrix of the first point cloud data to be calculated (i.e. Figure 2 Step S13 in (a).

[0108] According to the calculated second probability data, if it is determined that the second probability data is greater than the second preset probability threshold, after determining that the matching state of the valid point cloud data corresponding to the second probability data is a correct matching state, the valid point cloud data with a correct matching state is determined as the second point cloud data to be calculated, and the second probability data is based on the expression and expressions Calculate the Kalman gain and the innovation vector of the first point cloud data to be calculated under the uniform acceleration motion model.

[0109]

[0110] K t (K) = P(KK-1)H T (S t (K)) -1 ----expression

[0111] Where m represents the number of valid point clouds in the target confirmation matrix, V t It means that after calculating the comprehensive covariance matrix of all observations, the new information vector of the tth first point cloud data to be calculated, K t (K) represents the Kalman gain of the tth first point cloud data to be calculated.

[0112] At this time, each first point cloud data is traversed, and if the matching status of the traversed first point cloud data is a correct matching status, the first point cloud data to be calculated is based on the expression In combination with the Kalman gain and the new information vector, the motion state prediction value and the state covariance matrix prediction value obtained in step S12 are updated.

[0113] X t (K) = X t (KK-1)+K t (K)V t ----expression

[0114] Among them, X t (K) is the predicted value of the next motion state after update at time K, P t (K) is the predicted value of the next state covariance matrix after update at time K.

[0115] For ease of understanding, this embodiment provides the following scenario.

[0116] In this scenario, there are N motor vehicle targets, and the target set is T N ={1, 2, ..., N}, and there are N point cloud data Z to be matched at the current moment N ={1, 2, ..., N}, assuming that each motor vehicle target is associated with a target point cloud data, and the motor vehicle target associated with each target point cloud data is different.

[0117] At this time, the traditional joint probability data association algorithm is as follows:

[0118] The dimension of the initial confirmation matrix is: N×(N+1)

[0119] The number of interconnect matrices obtained by splitting the above initial confirmation matrix: 2 N

[0120] The dimension of the above interconnection matrix is: N×(N+1)

[0121] The computing resources occupied by the joint events corresponding to the interconnection matrix:

[0122]

[0123] The joint probability data association algorithm based on this embodiment is as follows:

[0124] The dimension of the initial confirmation matrix is: N×(N+1)

[0125] The number of target confirmation matrices obtained by splitting the above initial confirmation matrix: N

[0126] The dimension of the above target confirmation matrix is: 1×2

[0127] Based on the above objectives, confirm the number of interconnect matrices obtained by matrix splitting:

[0128] The dimensions of the above interconnect matrix: 1×2

[0129] The computing resources occupied by the joint events corresponding to the interconnection matrix:

[0130]

[0131] By obtaining the ratio between the computing resources obtained based on the traditional joint probabilistic data association algorithm and the computing resources obtained based on the joint probabilistic data association algorithm of this embodiment, it can be obtained that:

[0132]

[0133] According to the above ratio, when there is one maneuvering target vehicle, that is, N=1, the ratio between the computing resources obtained based on the traditional joint probability data association algorithm and the computing resources obtained based on the joint probability data association algorithm of this embodiment is M=1, that is, there is no difference between the computing resources occupied by the traditional joint probability data association algorithm and the computing resources occupied by the joint probability data association algorithm of this embodiment. However, when there are ten maneuvering target vehicles, that is, N=10, the ratio between the computing resources obtained based on the traditional joint probability data association algorithm and the computing resources obtained based on the joint probability data association algorithm of this embodiment is M=28160, that is, the computing resources required by the joint probability data association algorithm of this embodiment only need to occupy the computing resources required by the traditional joint probability data association algorithm. The computational resources of the joint probability data management algorithm are greatly reduced.

[0134] In this embodiment, a sliding window is set to perform multi-frame accumulation processing on the radar signal, and the position of the radar signal connected at the previous moment is compensated on the sliding window according to its motion state, thereby avoiding the missing of the radar signal at individual moments caused by the flickering problem of the radar signal. At the same time, it is proposed to extract valid point cloud data from the point cloud to be matched through an association gate, establish an initial confirmation matrix through the association relationship, and group the initial confirmation matrix, so as to partition the scenes corresponding to each radar signal to achieve the reduction of the dimension of the initial confirmation matrix, obtain multiple low-dimensional target confirmation matrices, and perform joint probability data association based on the low-dimensional target confirmation matrix, thereby reducing the number of interconnected matrices obtained by splitting the initial target confirmation matrix, and changing the exponential growth relationship between the number of joint events corresponding to the original interconnected matrix and the number of target point cloud data into an approximately linear relationship, which greatly reduces the computing resources of the joint probability data association algorithm.

[0135] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 3 Combination Figure 2 , step S10, the multi-target tracking method of the millimeter wave radar further includes steps S13 to S14:

[0136] Step S13, setting a plurality of the sliding windows, storing the plurality of radar signals outputted sequentially by the millimeter-wave radar into corresponding sliding windows, and extracting the radar signals stored in the plurality of sliding windows when a sliding window operation is detected to obtain a plurality of frames of point cloud data.

[0137] In order to reduce random fluctuations of radar signals and thus more clearly show the true trend of radar signals, this embodiment sets multiple sliding windows for storing radar signals.

[0138] Reference Figure 2As shown, F1, F2, ..., Fn-2, Fn-1, Fn, Fn+1, and Fn+2 are radar signals, and the radar signals Fn-2, Fn-1, Fn, Fn+1, and Fn+2 are stored in corresponding sliding windows respectively. If the next moment is entered at this time, and a new radar signal Fn-3 is connected at the next moment, the last frame of data Fn+2 is deleted, and the new radar signal Fn-3 is stored at the position where the radar signal Fn-2 is stored, and the radar signals Fn-2, Fn-1, Fn, and Fn+1 are moved backward one sliding window in sequence for The radar signals on the last few sliding windows are stored, and the position compensation is performed on the radar signals Fn-2, Fn-1, Fn, and Fn+1. Specifically, according to the time intervals between the radar signals Fn-2, Fn-1, Fn, and Fn+1 and the new radar signal Fn-3 and the motion states of the point cloud data stored in the radar signals Fn-2, Fn-1, Fn, and Fn+1, the positions of the point clouds stored in the radar signals Fn-2, Fn-1, Fn, and Fn+1 are adjusted. After the adjustment is completed, it is determined that a sliding window operation is completed (i.e. Figure 2 Step S1 in ).

[0139] Before extracting the radar signal, in order to ensure the validity of the extracted radar signal, each radar signal on the sliding window will be filtered for abnormal points according to the attributes such as SNRS, speed, distance, RCS, etc., and then the adjacent multi-frame radar signals will be accumulated and extracted to obtain multi-frame point cloud data (i.e. Figure 2 Step S2 in ).

[0140] Step S14, performing density-based clustering processing on the multi-frame point cloud data to obtain the point cloud set to be matched.

[0141] Each point cloud data on the extracted multi-frame point cloud data is traversed, and after marking the traversed point cloud data as visited, the domain point cloud set of each point cloud data is found. If the number of neighborhood points is less than the minimum number of neighborhood points, the point cloud data is marked as a noise point. If the number of neighborhood points is greater than the minimum number of neighborhood points, a new cluster is created, and the point cloud data and its neighborhood point cloud are added to the cluster, and then the cluster is expanded. For each point cloud in the neighborhood point cloud set, if the point cloud data has not been visited, the point cloud data is traversed and marked as visited, and then the neighborhood point cloud set of the point cloud data is found. If the number of neighborhood points is greater than or equal to the minimum number of neighborhood points, the neighborhood point set of the point cloud number is added to the neighborhood point set of the cluster. If the number of neighborhood points is greater than the minimum number of neighborhood points, a new cluster is created, and so on, until all point cloud data are traversed and clustered to obtain multiple point cloud clusters (i.e. Figure 2Step S3 in ), each point cloud cluster is regarded as a point cloud set to be matched (i.e. Figure 2 In step S4), the attribute information of the point cloud set to be matched is calculated by integrating the attribute information of all point cloud data in the point cloud cluster.

[0142] In this embodiment, a sliding window is set to perform multi-frame accumulation processing on the radar signal, and the position of the radar signal connected at the previous moment is compensated on the sliding window according to its motion state, thereby avoiding the missing of radar signals at individual moments caused by the flickering problem of the radar signal. At the same time, density-based clustering processing is used to extract the point cloud set to be matched, so as to obtain point cloud data with similar characteristics, so as to facilitate the processing of point cloud data with similar characteristics to be matched in subsequent steps.

[0143] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 4 Combination Figure 2 After step S43, the multi-target tracking method of the millimeter wave radar further includes steps A11 to A13:

[0144] Step A11: determine the valid point cloud data other than the second point cloud data to be calculated as the point cloud data to be determined whose matching status is an incorrect matching status.

[0145] Step A12, determining whether the point cloud data to be determined meets the preset creation conditions.

[0146] Step A13: If the point cloud data to be determined meets the preset creation condition, the point cloud data to be determined is determined as the first next target point cloud data of the next target tracking process (i.e. Figure 2 , step S14 in the process.

[0147] It should be noted that steps S10 to S40 are continuously executed in a loop based on the multi-frame tracking target output by the sliding window operation. The multi-target tracking operation performed based on the acquired multi-frame tracking target is an operation for tracking the target motion generated at the current moment. Because the association probability of the valid point cloud data at different moments and the first point cloud data is not the same, the valid point cloud data whose matching state is the correct matching state is only the current moment. The point cloud data to be judged whose matching state is judged to be the wrong matching state needs to be judged whether it meets the target tracking conditions of the multi-target tracking operation at the next moment, that is, the preset creation conditions, and judges the point cloud properties of the point cloud data to be judged, such as whether the speed meets the speed requirements of the preset creation conditions corresponding to the next moment. If yes, the point cloud data to be judged is input into the multi-target tracking operation at the next moment, that is, the point cloud data to be judged is determined as the first next target point cloud data of the next target tracking process, so as to avoid the situation where the continuity of multi-target tracking is reduced due to the lack of target point cloud data.

[0148] After step S40, the multi-target tracking method of the millimeter wave radar further includes steps S51-S52:

[0149] Step S51, traverse the tracks and point cloud attributes corresponding to each of the first point cloud data, extract the first point cloud data with a track status of a dead track from each of the tracks, and extract the first point cloud data with an attribute status of an abnormal attribute from each of the point cloud attributes, and then remove the first point cloud data with a track status of the dead track and / or the first point cloud data with an attribute status of an abnormal attribute to obtain the second next target point cloud data (i.e. Figure 2 .

[0150] Step S52: The first next target point cloud data and the second next target point cloud data are transferred to the next target tracking process to calculate each of the tracked targets (i.e. Figure 2 , step S16 in the process).

[0151] In this embodiment, by traversing the tracks and point cloud attributes corresponding to each first point cloud data, the first point cloud data that cannot be input into the multi-target tracking operation at the next moment is extracted for multi-target tracking. Specifically, each first point cloud data is traversed to determine whether the track corresponding to it overlaps with the track of other first point cloud data. If there is overlapping information, the track state of the track is determined as a dead track, and it is determined whether the corresponding track is a track split from other vehicles. If so, the track state of the track is determined as a dead track, and according to the point cloud attributes, it is determined whether the corresponding first point cloud data has a track abnormality, such as speeding. If so, after determining that the attribute state of the point cloud data is an abnormal attribute, the first point cloud data with a track state of a dead track and / or an attribute state of an abnormal attribute is eliminated, and the first point cloud data that is not eliminated is input as the second next target point cloud data into the multi-target tracking operation at the next moment.

[0152] In this implementation, by screening the first point cloud data at the current moment, the next target point cloud data that can be input into the multi-target tracking operation at the next moment is obtained, thereby ensuring the continuity of the multi-target tracking.

[0153] The present application provides a multi-target tracking device of a millimeter-wave radar, and the multi-target tracking device of the millimeter-wave radar includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-target tracking method of the millimeter-wave radar in the above-mentioned embodiment one.

[0154] Reference below Figure 5 , which shows a schematic diagram of the structure of a multi-target tracking device suitable for implementing the millimeter-wave radar of the embodiment of the present application. The multi-target tracking device of the millimeter-wave radar in the embodiment of the present application may include but is not limited to mobile terminals such as laptop computers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The multi-target tracking device of the millimeter-wave radar shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0155] like Figure 5As shown, the multi-target tracking device of the millimeter wave radar may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the multi-target tracking device of the millimeter wave radar are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the multi-target tracking device of the millimeter-wave radar to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a multi-target tracking device of the millimeter-wave radar with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0156] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0157] The multi-target tracking device of the millimeter-wave radar provided by the present application adopts the multi-target tracking method of the millimeter-wave radar in the above-mentioned embodiment, which can solve the technical problem that the multi-target tracking based on the joint probability data association algorithm currently has too much computing resources. Compared with the prior art, the beneficial effects of the multi-target tracking device of the millimeter-wave radar provided by the present application are the same as the beneficial effects of the multi-target tracking method of the millimeter-wave radar provided by the above-mentioned embodiment, and the other technical features of the multi-target tracking device of the millimeter-wave radar are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0158] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0159] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0160] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the multi-target tracking method of the millimeter-wave radar in the above-mentioned embodiment.

[0161] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0162] The computer-readable storage medium may be included in the multi-target tracking device of the millimeter-wave radar; or may exist independently without being assembled into the multi-target tracking device of the millimeter-wave radar.

[0163] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the multi-target tracking device of the millimeter-wave radar, the multi-target tracking device of the millimeter-wave radar: by setting multiple sliding windows, multi-frame accumulation processing and density-based clustering processing are performed on the accessed radar signals to obtain a point cloud set to be matched, and by traversing the point cloud set to be matched, a predicted point cloud set is calculated; an association gate is established based on the association relationship between the predicted point cloud set and the point cloud set to be matched; based on the association gate, valid point cloud data in the point cloud set to be matched and first point cloud data corresponding to the valid point cloud data in the predicted point cloud set are obtained, and an initial confirmation matrix is ​​established based on the association relationship between the valid point cloud data and the first point cloud data; multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix are obtained, and the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index are grouped into one group to obtain multiple groups, and a target confirmation matrix belonging to the group is established, and joint probability data association is performed based on the target confirmation matrix of each group, wherein the predicted point cloud set includes a plurality of predicted point cloud data.

[0164] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0165] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0166] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0167] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the multi-target tracking method of the above-mentioned millimeter-wave radar, and can solve the technical problem that the multi-target tracking based on the joint probability data association algorithm currently has too much computing resources. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the multi-target tracking method of the millimeter-wave radar provided by the above-mentioned embodiment, and will not be repeated here.

[0168] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A multi-target tracking method of millimeter wave radar, characterized in that: The multi-target tracking method of the millimeter wave radar comprises: By setting multiple sliding windows, the radar signals are processed with multi-frame accumulation and density-based clustering to obtain the point cloud set to be matched, and the predicted point cloud set is calculated by traversing the target point cloud set at the previous moment; Establishing an association gate according to the association relationship between the predicted point cloud set and the to-be-matched point cloud set; Based on the association gate, valid point cloud data in the to-be-matched point cloud set and first point cloud data corresponding to the valid point cloud data in the predicted point cloud set are acquired, and an initial confirmation matrix is ​​established according to the association relationship between the valid point cloud data and the first point cloud data; Acquire multiple predicted point cloud indexes associated with each valid point cloud data in the initial confirmation matrix, group the valid point cloud data with the same predicted point cloud index and the predicted point cloud data pointed to by the predicted point cloud index into one group, obtain multiple groups, and establish a target confirmation matrix belonging to the groups, perform joint probability data association based on the target confirmation matrices of each group, wherein the predicted point cloud set includes several predicted point cloud data.

2. The multi-target tracking method of millimeter wave radar as claimed in claim 1, characterized in that: The step of setting multiple sliding windows, performing multi-frame accumulation processing and density-based clustering processing on the accessed radar signals, and obtaining a point cloud set to be matched includes: Setting a plurality of sliding windows, storing a plurality of radar signals sequentially output by the millimeter-wave radar in corresponding sliding windows, and extracting radar signals stored in a plurality of frames of the sliding windows when a sliding window operation is detected to be completed, to obtain a plurality of frames of point cloud data; The multi-frame point cloud data is subjected to density-based clustering processing to obtain the point cloud set to be matched.

3. The multi-target tracking method of millimeter wave radar as claimed in claim 2, characterized in that: The millimeter wave radar includes a plurality of target trackers, one of the target trackers is used to track a point cloud data, and the step of traversing the point cloud set to be matched and calculating the predicted point cloud set includes: Initializing the target tracker; After traversing the target track corresponding to each target point cloud data in the target point cloud set at the previous moment, the motion state prediction value of each target point cloud data at the current moment and the state covariance matrix prediction value at the current moment are calculated through the uniform acceleration motion model to obtain the predicted point cloud set.

4. The multi-target tracking method of millimeter wave radar as claimed in claim 3, characterized in that: The to-be-matched point cloud set includes a plurality of to-be-matched point cloud data, and the step of establishing an association gate according to the association relationship between the predicted point cloud set and the to-be-matched point cloud set includes: Traversing each of the to-be-matched point cloud data, and traversing the predicted point cloud data corresponding to the target point cloud data based on the traversed to-be-matched point cloud data, to obtain target predicted point cloud data corresponding to each of the target point cloud data; According to the point cloud data to be matched, the association gate is established on each of the target prediction point cloud data.

5. The multi-target tracking method of millimeter wave radar as claimed in claim 4, characterized in that: The step of acquiring valid point cloud data in the to-be-matched point cloud set based on the association gate comprises: By means of the association gate, determining whether the target point cloud data falls within the association range of the association gate; If the target point cloud data falls within the associated range, the to-be-matched point cloud data corresponding to the target point cloud data falling within the associated range is extracted as the valid point cloud data.

6. The multi-target tracking method of millimeter wave radar as claimed in claim 5, characterized in that: The step of performing joint probability data association according to the target confirmation matrix of each group comprises: Traversing each of the target confirmation matrices, and splitting each of the target confirmation matrices to obtain a plurality of interconnection matrices, wherein each of the interconnection matrices represents a feasible joint event; Calculating the probability of the feasible joint event according to the interconnection matrix to obtain first probability data, and calculating the probability that the valid point cloud data included in the interconnection matrix belongs to the first point cloud data to obtain second probability data; According to the comparison relationship between the second probability data and the second preset probability threshold, if the second probability data is greater than the second preset probability threshold, the matching state of the valid point cloud data corresponding to the second probability data is marked as a correct matching state, wherein the first point cloud data in the correct matching state is determined as the first point cloud data to be calculated, and the valid point cloud data in the correct matching state is determined as the second point cloud data to be calculated; Calculate the Kalman gain and the new information vector of the first point cloud data to be calculated under the uniform acceleration motion model, and update the motion state prediction value and the state covariance matrix prediction value according to the Kalman gain and the new information vector to obtain the motion state and the posterior state covariance of the first point cloud data to be calculated.

7. The multi-target tracking method of millimeter wave radar as claimed in claim 6, characterized in that: After the step of marking the matching state of the valid point cloud data corresponding to the second probability data as a correct matching state if the second probability data is greater than the second preset probability threshold, the method further includes: Determine the valid point cloud data other than the second point cloud data to be calculated as the point cloud data to be determined whose matching status is an incorrect matching status; Determine whether the point cloud data to be determined meets the preset creation conditions; If the point cloud data to be determined meets the preset creation condition, the point cloud data to be determined is determined as the first next target point cloud data of the next target tracking process.

8. The multi-target tracking method of millimeter wave radar as claimed in claim 7, characterized in that: After the step of performing joint probability data association according to the target confirmation matrix, the method further comprises: Traversing the tracks and point cloud attributes corresponding to each of the first point cloud data, extracting the first point cloud data with a track status of a death track from each of the tracks, and extracting the first point cloud data with an attribute status of an abnormal attribute from each of the point cloud attributes, then eliminating the first point cloud data with a track status of the death track and / or the first point cloud data with an attribute status of an abnormal attribute, to obtain the second next target point cloud data; The first next target point cloud data and the second next target point cloud data are passed into the next target tracking process.

9. A millimeter wave radar multi-target tracking device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the multi-target tracking method of the millimeter-wave radar according to any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the multi-target tracking method of the millimeter-wave radar according to any one of claims 1 to 8 are implemented.

Images